I've had to say this in enough threads that I felt it was worth repeating in a diary. I want to talk about Zogby's Iowa tracking poll, and today's results (as well as Zogby's comments).
Specifically, I'd like people to treat a tracking poll as it should, and not simply believe what they want to believe.
We all know Zogby's latest numbers: Dean 24 (- 4), Gephardt 21 (-2), Kerry 21 (+4).
Big swings, for sure. What does Zogby say: "This is officially a 3-way race. Kerry is surging. He actually led the pack on Monday with 25% for the day. Dean had his worst single day with only 18%.".
That's a darn big jump. Why, by that measure Dean lost almost 30% of his support in a single day, and Kerry increased his support by almost 60%.
Well, not really. First off, when Zogby talks about "Monday's sample" he's referring to referring to the 167 people his firm polled on Monday. (They poll 501 people over the course of three days).
Each daily subsample has an MOE of around 7.75%, if I did the math right. And, of course, taking samples daily increases the risk of getting a skewed sample.
That's where three day averages come in. Because three days worth of polling are averaged together, the sample size is larger and the swings caused by the small size of the sub-sample are damped down. Nonetheless, tracking polls will fluctuate a lot.
So looking at the numbers, we know Dean had an 18 on Monday, and Kerry had a 25.
We don't know Sunday's numbers, or Tuesday's numbers, but we can calculate the average of those two numbers. Dean's Sunday and Monday figures had to average 27, and Kerry's had to average 19.
Here's the fun part: If Monday's numbers were real, and not the result of the usual stastical fluke, then Tuesdays should have been similiar. (Though you would have thought that Zogby would have mentioned that, wouldn't you? I mean, if Edwards hit 25 two days in a row....).
But that would require both Edwards and Dean to have numbers on Sunday that were significantly out of whack with their average...yet Zogby didn't mention those.
Zogby didn't mention a continuation of the "Kerry surge" on Tuesday, nor did he comment on extra-high Dean numbers on Sunday.
Actually, it gets even better because Dean's moving average yesterday was a 28, which meant the he had to be polling around 33% Sunday and Saturday...yet Zogby didn't remark on a fifteen point Dean fall, almost half his support?
Kerry, of course, would have averaged 13 Sunday and Saturday, which meant Monday was a 25% jump.
Now, this isn't to say that Kerry didn't rise, or Dean didn't fall. I wouldn't be surprised at either movement. But right now, Monday's take looks like an outlier. It's hard to see a candidate losing 30% of his support, or gaining half again as many supporters in a single day...and then failing to hold those gains or losses the next day.
In short: I would be very skeptical of Monday's return, and thus the numbers released yesterday, today, and tommorow for Zogby's Iowa tracking poll. Friday's numbers should verify if this is a real trend, or just a fluke, or a little of both.
It wouldn't surprise me at all to see Kerry rising some and Dean falling a bit. Nonetheless, these numbers feel far more like a statistical fluke than the beginnings of a trend.
Perhaps I'm wrong. Friday will tell. But don't be surprised to see a big jump for Dean, and a big fall for Kerry, when the three day average for Tuesday, Wednesday, and Thursday is released.